Deep Learning Workshop with Julia

This session will provide an in-depth view into how to create and deploy practical deep learning models with Julia. We hope to cater to a wide range of interests, from the beginner to the expert machine learner.

We will start with a brief exercise in creating a neural network by hand. High level Julia code makes that feasible, and that will provide the necessary background as we move into higher level abstractions.

The rest of the session will be based on Flux.jl, the elegant machine learning stack. We will begin by creating simple models, and learn the api's used to create, train and validate models, as well as ways to create the necessary feature encodings.

We will then go on to create more complex, practical models, with examples in natural language processing and image processing. We will discuss ways to train models on the GPU, port pre-trained models from other frameworks, and see how to export and import models for moving them to production systems.

The session will be conducted by Avik Sengupta (Julia Computing) and Pontus Stenetorp (UCL). We will use Juliabox to demonstrate the code, please bring your laptop to follow along.

Some knowledge of Julia will be helpful in getting the most of this session. But if you haven't learnt Julia yet, fear not. Jane Herriman's video tutorial on youtube is the best way to get started, and will be sufficient background for this session. Watch it here: https://www.youtube.com/watch?v=zJ_d3oWuogw

Registration, Networking and Pizzas: 6:15 PM
Talks Start: 6:45 PM
Talks End: 8:45 PM